Anthropic Releases Claude Fable 5.1, Cuts Agentic Workload Pricing Up to 45%

Anthropic has released Claude Fable 5.1, an upgrade to its top-tier Fable 5 model that debuted in June. The company says the new model improves performance on coding, knowledge work, and long-running problem-solving tasks while cutting costs on agentic workloads by up to 45%.

Where Fable sits in Claude's lineup

Fable is Anthropic's most capable model class, sitting above Haiku, Sonnet, and Opus in the company's tier structure. Anthropic describes Fable as "Mythos-class," a designation the company introduced with Fable 5 in June 2026.

Alongside Fable 5.1, Anthropic is releasing Mythos 5.1 for users with Mythos access. According to Anthropic, "Claude Fable 5.1 and Claude Mythos 5.1 are the same model, but with different levels of safeguards." Fable 5.1 is generally available; Mythos 5.1 is restricted to "trusted access programs," with safeguards specifically built for cybersecurity and life sciences work.

Performance claims

Anthropic claims Fable 5.1 "sets a new standard on coding, knowledge work, and long-running problem-solving tasks" and that it "achieves similar or better results than Fable 5 at a much lower cost" when run at low or medium effort settings. The company did not publish specific benchmark scores in its announcement.

As an example of the model's capability, Anthropic cited testing by investment firm Millennium, in which Fable 5.1 reportedly identified the root cause of a rare crash in Millennium's internal systems — one that had gone unexplained by the firm's engineers, and by other AI models, for several years. This claim comes from Anthropic's announcement and has not been independently verified.

Anthropic also states that Fable 5.1 "avoids shortcuts that result in poorer-quality work" and is capable of fixing root causes of software issues rather than surface-level symptoms — a claim consistent with the Millennium example but otherwise unsubstantiated by public benchmark data.

Pricing changes

Anthropic is not changing base token pricing for Fable 5.1 relative to Fable 5. Instead, the company is cutting the price of cache reads — instances where the model reuses previously processed context — for any usage billed by token, including API access.

According to Anthropic, this cache-read discount translates to roughly 25% lower total cost for typical workloads. For highly agentic work, where cache reuse is more frequent, Anthropic says savings can reach approximately 45%. Anthropic has not disclosed specific per-token dollar figures for either the base rate or the discounted cache-read rate; pricing not yet disclosed in absolute terms.

No context window figure was included in Anthropic's announcement.

What this means

The headline change here isn't raw capability — Anthropic offers no benchmark scores to substantiate its "new standard" claim — it's cost structure. By discounting cache reads rather than base rates, Anthropic is specifically targeting long-running agentic workflows and coding assistants that repeatedly reprocess large contexts, which is where cache reuse adds up fastest. The Mythos 5.1 gating for cybersecurity and life sciences use cases also signals Anthropic is treating its most capable model tier as dual-use technology requiring separate access controls, a pattern likely to continue as frontier models grow more capable on offensive security and biological research tasks. Until independent benchmarks surface, buyers should treat Anthropic's performance claims — including the Millennium anecdote — as company-reported rather than verified.